Dynamic Compensation of Harmonic, Sag and Interruption Using Unified Neural Network Controller with Levenberg-Marquardt Algorithm in Series Active Power Filter

This paper proposes a unified neural network controller in development of series active power filter. Three major power quality problems considered for this work are harmonics, sags, and interruptions, and dealing with all of them really needs effective unified algorithm which could adapt with variation of signals. Therefore, Levenberg-Marquardt (L-M) back propagation algorithm is introduced in this controller which helps the proposed controller to work properly with dynamic responses with appropriate achievement of convergence. Simulation results clearly show high performance of the proposed controller with series APF system. Its ability of adaptive estimating enhances the response time of the system to only one cycle. It performs well under various testing conditions with significant reduction of harmonics, correction of sags, and fast response to interruptions.